Privacy
Last updated 7 September 2026
You are handing us other people's documents. This page says exactly what happens to them, in the order that matters to you, without the hedging that usually fills a policy like this.
The short version
- Your documents are never used to train anything. Not our models, not anyone else's.
- Your documents are never sold, rented or shared. Not with advertisers, not with data brokers, not with partners.
- No unauthorised person reads them. Nobody at dextract opens your documents to browse them. The narrow exceptions are named below and every one of them is logged.
- Everything gets deleted. Trial documents and results are permanently removed after 30 days. Delete one yourself and it goes sooner, and you can delete your whole account from Settings. The one thing we keep is a count of trial usage that cannot be traced back to you.
What we collect
Three things and nothing beyond them.
| What | Where it comes from |
|---|---|
| Your account | Your name, email address and profile picture, from the Google account you sign in with. We never see or store a password. |
| Your documents | The files you upload and the results we produce from them. |
| An access record | Who did what and when, with the IP address and browser the request came from. It contains no document content. |
We do not run advertising trackers, analytics scripts or third party pixels on this site. There is nothing on these pages watching you read them.
What we do with it
We use your documents for exactly one purpose: to read them and give you the results, on the request you made. That processing is automated from end to end.
We use your account details to sign you in, keep your work separate from every other customer and reply when you write to us. We use the access record to investigate problems, to answer your own security questions about who touched what and to keep the trial allowance honest.
What we never do
- No training. Your documents are never used to train, tune, evaluate or improve any model, ours or anyone else's. dextract builds and owns the AI models that read your documents and they are trained before your file ever arrives. What we learn from your documents is your results and those are yours.
- No selling or sharing. We do not sell, rent, licence or trade your documents, your results or your account details. There is no arrangement under which anyone pays us for access to them.
- No third party AI service. Your documents are not sent to an outside model provider for processing. They are read by our own models, on machines we operate.
- No browsing. Nobody at dextract opens your documents out of curiosity, to build a demo, or to check how a customer is doing.
Who can see your documents
You and the people in your organisation. Access is enforced in the database itself, so a request carrying one organisation's session cannot read another's records even if something above it goes wrong.
There are two narrow cases where someone at dextract can reach a stored document:
- You ask us to. If you report a document that read badly and ask us to look, a named engineer opens that specific document to fix it. We do not go looking without that request.
- The law compels us. If we receive a valid legal order we will comply and we will tell you unless we are prohibited from doing so.
Both leave an entry in the access record, which you can ask us for at any time.
Where it runs
The models that read your documents run on hardware we own and operate. The application, the database and the stored files run on infrastructure we rent in the United States, from providers who hold the data on our behalf under contract and have no right to use it:
| Provider | What it holds |
|---|---|
| Sign in only. Google confirms who you are and never receives your documents. | |
| Cloudflare | Object storage for your uploaded files and results. |
| Railway | Application hosting and the database of accounts and job records. |
If you need your documents to stay inside your own environment, that is what our on-premise deployment is for. Nothing leaves your network and this page stops applying because we never hold anything.
How long we keep it
Trial documents and their results are permanently removed 30 days after upload. This is automatic. You do not have to ask.
Delete a document yourself and it stops being readable immediately, then the stored bytes are permanently removed. It is not recoverable, by you or by us.
Two things survive a deletion on purpose. The filename stays on your dashboard marked as deleted, so a document does not silently vanish from your own records and the access record of the deletion is kept, so there is a trail of what was removed and when. Neither contains any content from the document.
Your account details are kept while your account exists. Ask us to close it and they go too.
Security
- Everything is encrypted in transit. Stored files sit in encrypted object storage.
- Sign in is by Google, so we never hold a password to lose.
- Separation between organisations is enforced at the database, not just in application code.
- Every meaningful action is written to an access record that no ordinary user can edit.
Cookies
One cookie, holding your signed in session. It is HTTP only and same site. There are no advertising cookies, no tracking cookies and no third party cookies, so there is no banner to click through.
Your requests
Write to us and we will act on any of the following, at no cost:
- A copy of everything we hold about you.
- Deletion of your documents, your results, or your whole account.
- Correction of your account details.
- The access record showing who touched your documents.
The second one you can do yourself, right now: Settings, then Delete this workspace. It removes every document, every extraction and your account, and it erases the files from storage immediately rather than waiting out the retention window. Nobody here has to be involved and nothing is queued for later.
For anything else, write to us. We answer within 30 days, usually the same week. Reach us through the contact page.
Changes
If we change this policy in a way that affects what happens to your documents, we will tell account holders by email before it takes effect rather than quietly updating the date at the top.